CN109587713B - Network index prediction method and device based on ARIMA model and storage medium - Google Patents
Network index prediction method and device based on ARIMA model and storage medium Download PDFInfo
- Publication number
- CN109587713B CN109587713B CN201811485502.2A CN201811485502A CN109587713B CN 109587713 B CN109587713 B CN 109587713B CN 201811485502 A CN201811485502 A CN 201811485502A CN 109587713 B CN109587713 B CN 109587713B
- Authority
- CN
- China
- Prior art keywords
- arima model
- sequence
- prediction
- index
- determining
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04W—WIRELESS COMMUNICATION NETWORKS
- H04W24/00—Supervisory, monitoring or testing arrangements
- H04W24/02—Arrangements for optimising operational condition
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04W—WIRELESS COMMUNICATION NETWORKS
- H04W16/00—Network planning, e.g. coverage or traffic planning tools; Network deployment, e.g. resource partitioning or cells structures
- H04W16/22—Traffic simulation tools or models
Landscapes
- Engineering & Computer Science (AREA)
- Computer Networks & Wireless Communication (AREA)
- Signal Processing (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
- Data Exchanges In Wide-Area Networks (AREA)
Abstract
The invention discloses a network index prediction method, a device and a storage medium based on an ARIMA model, wherein the method comprises the following steps: acquiring index data of an index variable to be predicted in a certain time period as a training data set; taking the training data set after preprocessing as input, and constructing an ARIMA model for predicting network indexes; and inputting the set future time length of the index variable to be predicted and the selected stable sequence which passes the d-order difference into the ARIMA model together, and calculating to obtain a target predicted value. According to the invention, through counting the change rule of the user on the wireless network demand, the change sequence value of a certain index in a certain time in the future can be predicted, and more effective data reference is provided for optimizing wireless network resource allocation and performance optimization.
Description
Technical Field
The invention relates to the technical field of data mining, in particular to a network index prediction method and device based on an ARIMA model and a storage medium.
Background
With the rapid development of communication information technology and the large-scale popularization of wireless networks, the demands of huge user quantity on the networks make the traditional operation and maintenance means have to seek a more efficient data decision method. How to quantitatively predict the network use condition of the user in advance and accurately, avoid or reduce the occurrence probability of network blockage, improve the resource allocation efficiency, and is one of the key contents of daily operation and maintenance optimization of communication operators.
At present, methods for predicting and discovering network communication problem conditions in advance are insufficient, problem analysis after the problem is usually based on simple one-dimensional linear analysis among indexes, and wireless network requirements in a future period cannot be predicted and fed back in advance by effectively integrating time sequence change rule data.
In the research and practice of the prior art, the inventor of the present invention finds that the existing network communication index analysis method mainly faces the following problems:
1) the conventional methods such as a linear prediction method and a neural network have poor practicability, the prediction result is unstable under a small sample, and the situation of a time sequence change rule cannot be well compatible;
2) the periodic change of the index is easy to ignore.
Disclosure of Invention
The technical problem to be solved by the invention is to provide a network index prediction method, a device and a storage medium based on an ARIMA model, which can predict the change sequence value of a certain index in a certain time in the future by counting the change rule of the wireless network demand of a user, and provide more effective data reference for optimizing wireless network resource allocation and performance optimization.
To solve the above problem, an embodiment of the present invention provides a network index prediction method based on an ARIMA model, including:
acquiring index data of an index variable to be predicted in a certain time period as a training data set;
taking the training data set after preprocessing as input, and constructing an ARIMA model for predicting network indexes;
and inputting the set future time length of the index variable to be predicted and the selected stable sequence which passes the d-order difference into the ARIMA model together, and calculating to obtain a target predicted value.
Further, the training data set comprises an index data sequence Xtrain={x1,x2,…xLAnd time series L ═ L1,L2,…LL};
The training data set is preprocessed by decomposing the training data set into a trend part, a period part and a residual part; in particular, the method comprises the following steps of,
Xtrain={T1,T2,…TL}+{S1,S2,…SL}+{R1,R2,…RL};
xi=Ti+Si+Ri,i=1,2,…,L;
wherein, Ti、Si、RiRespectively is a trend component, a period component and an error term which are decomposed from the index data sequence.
Further, the building of the ARIMA model for predicting the network index refers to determining parameters p, d, q and other related parameters of the ARIMA model;
determining parameters p, d, q specifically includes:
performing stationarity test and d-order difference processing on the trend component sequence obtained by preprocessing until a stationarity sequence is obtained, and determining a parameter d;
after the stable sequence is obtained, determining a parameter p and a parameter q by respectively using a PACF method and an ACF method;
determining other related parameters specifically comprises:
after determining the parameters p, d, q, the prediction formula of the ARIMA model is expressed as:
where μ is a constant term, ρjIs the coefficient of an autoregressive process of order p, epsiloniIs an error constant, θjError term coefficients for a moving average process of order q;
after the periodic component sequence obtained by preprocessing is added, the prediction formula of the ARIMA model is expressed as,
further, the network index prediction method based on the ARIMA model further includes:
and superposing the target predicted value and the periodic component sequence to obtain an actual predicted value.
Further, the network index prediction method based on the ARIMA model further includes:
and carrying out error calculation on the target predicted value and the actual predicted value to obtain a standard deviation and a relative error.
Another embodiment of the present invention further provides an ARIMA model-based network index prediction apparatus, including:
the acquisition module is used for acquiring index data of an index variable to be predicted in a certain time period as a training data set;
the model construction module is used for taking the training data set after preprocessing as input to construct an ARIMA model for predicting network indexes;
and the prediction module is used for inputting the set future time length of the index variable to be predicted and the selected stable sequence which passes the d-order difference into the ARIMA model together to calculate and obtain a target prediction value.
Further, the ARIMA model-based network index prediction apparatus further includes a preprocessing module configured to decompose the training data set into a trend part, a period part, and a residual part; in particular, the method comprises the following steps of,
Xtrain={T1,T2,…TL}+{S1,S2,…SL}+{R1,R2,…RL};
xi=Ti+Si+Ri,i=1,2,…,L;
wherein the training data set comprises an index data sequence Xtrain={x1,x2,…xLAnd time series L ═ L1,L2,…LL};Ti、Si、RiRespectively is a trend component, a period component and an error term which are decomposed from the index data sequence.
Further, the ARIMA model-based network index prediction apparatus further includes:
and the prediction module is also used for superposing the target predicted value and the periodic component sequence to obtain an actual predicted value.
Further, the ARIMA model-based network index prediction apparatus further includes:
and the error analysis module is used for carrying out error calculation on the target predicted value and the actual predicted value to obtain a standard deviation and a relative error.
Yet another embodiment of the present invention further provides a computer-readable storage medium comprising a stored computer program, wherein the computer program, when running, controls a device on which the computer-readable storage medium is located to perform the ARIMA model-based network index prediction method as described above.
The embodiment of the invention can effectively count the change rule of the user on the wireless network demand from the change of the historical data, effectively predict the change sequence value of a certain index in a certain time in the future, and provide more effective data reference for optimizing wireless network resource allocation and performance optimization.
Drawings
Fig. 1 is a schematic flow chart of a network index prediction method based on an ARIMA model according to an embodiment of the present invention;
fig. 2 is another schematic flow chart of a network index prediction method based on an ARIMA model according to an embodiment of the present invention;
fig. 3 is a schematic flow chart of a network index prediction method based on an ARIMA model according to an embodiment of the present invention;
fig. 4 is a schematic structural diagram of a network index prediction apparatus based on an ARIMA model according to an embodiment of the present invention;
fig. 5 is a schematic structural diagram of a network index prediction apparatus based on an ARIMA model according to an embodiment of the present invention.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
It should be noted that the step numbers presented herein are only for convenience of description and are not limited to the execution order of the steps.
Please refer to fig. 1-3. One embodiment of the invention provides a network index prediction method based on an ARIMA model, which comprises the following steps:
and S1, acquiring index data of the index variable to be predicted in a certain time period as a training data set.
Wherein the training data set comprises an index data sequence Xtrain={x1,x2,…xLAnd time series L ═ L1,L2,…LL};
In a specific embodiment, data within a certain historical time period (granularity of 1 hour/1 day) of the network communication index variable may be selected as a training data set of the model.
For example, a certain city is taken as a unit, time series historical data of a plurality of index variables collected by a base station of the city is selected, and if the change condition of the telephone traffic in a future week of a certain area needs to be predicted, the telephone traffic indexes with the granularity of the whole network, network elements or cells as the area granularity and the granularity of 1 hour as the selection time are taken as training data respectively.
And S2, constructing an ARIMA model for predicting network indexes by taking the preprocessed training data set as input.
The preprocessing of the training data set comprises decomposing the training data set into a trend part, a period part and a residual part; in particular, the method comprises the following steps of,
Xtrain={T1,T2,…TL}+{S1,S2,…SL}+{R1,R2,…RL};
xi=Ti+Si+Ri,i=1,2,…,L;
wherein, Ti、Si、RiRespectively is a trend component, a period component and an error term which are decomposed from the index data sequence.
In a specific embodiment, the periodic feature extraction is performed on a sequence with periodic features according to different time granularities, where a common algorithm STL (secure-Trend decoding procedure on Loess) in time sequence periodic rule computation decomposition is used, and data x at a certain time is extracted based on the LOESSiIs decomposed into trend component (T) and period componentA quantity (S) and an error term (R).
Note that the trend component (T)i) The method is characterized by comprising a trend part; periodic component (S)i) Embodied by a periodic portion; error term (R)i) Embodying the residual part.
The building of the ARIMA model for predicting the network index refers to determining parameters p, d, q and other related parameters of the ARIMA model.
Determining parameters p, d, q specifically includes:
performing stationarity test and d-order difference processing on the trend component sequence obtained by preprocessing until a stationarity sequence is obtained, and determining a parameter d;
after the stationary sequence is obtained, the parameter p and the parameter q are determined by the PACF method and the ACF method, respectively.
In a specific embodiment, after the step periodic rule processing, since the trend component sequence represents the variation in the original data sequence, because the trend component sequence T ═ T1,T2,…TLAnd fourthly, carrying out next model parameter training.
And (3) carrying out statistical testing on stationarity: adopting ADF (automatic-Dickey-filler test) method to determine T as { T ═ T1,T2,…TLThe stationarity of the sequence is checked by unit root, if ADF passes, i.e. the sequence T '═ T'1,T′2,…T′LThe sequences are plateau sequences.
If the ADF check fails, i.e., the sequence T is a non-stationary sequence (there is a root of a unit), the sequence needs to be first order differenced, i.e.,
Ti_diff={Ti2-Ti1,Ti3-Ti2,…Tin-Tin-1}。
generally, the trend component sequence satisfies the ADF inspection after passing through the first order difference, and reaches a smooth sequence. However, if the first-order difference still fails to pass the ADF inspection, the cyclic difference processing is performed until the sequence stationary T ' ═ T ' is obtained '1,T′2,…T′LAnd d (the order of the cyclic difference).
Determining a parameter p: to obtain T '═ T'1,T′2,…T′LAfter the smoothing sequence, for the autoregressive process ar (p), the Partial Autocorrelation Coefficient (PACF) method is used to represent:
Φ11=ρ1;
wherein, using phikjRepresents the jth regression coefficient, rho, in the k-th order autoregressive equationjIs an autocorrelation coefficient, is obtained by the least square method, phikj=Φk-1,j-ΦkkΦk-1,k-1J ═ 1,2,3, …, k-1; when k is less than or equal to p, phikkNot equal to 0; when k is>p is phikkApproaching 0, i.e. T'LAnd T'L-kThe partial autocorrelation coefficient of (a) is close to 0, which shows that the truncation characteristics of the partial autocorrelation function after the lag period p are effective, so that the order of the AR (p) process can be identified by using the characteristics to obtain a model parameter p.
Determining a parameter q: to obtain T '═ T'1,T′2,…T′LAfter the smoothing sequence, since the moving average method aims to effectively eliminate random fluctuation in prediction, for the moving average model process ma (q), the autocorrelation function (ACF) method is used to represent:
wherein, L is the sample capacity of the test sequence,is the sequence mean. According to the stable characteristic of the T' sequence, P can be judgedkIs a convergent sequence, i.e. when k>q is, PqApproaching 0 to determine the model parameters q.
Determining other related parameters specifically comprises:
after determining the parameters p, d, q, the prediction formula of the ARIMA model is expressed as:
where μ is a constant term, ρjIs the coefficient of an autoregressive process of order p, epsiloniIs an error constant, θjIs the error term coefficient of the moving average process of the order q.
In a specific embodiment, when there are multiple sets of candidate parameters p, d, and q that can be selected, the model parameters are selected according to the AIC and BIC values, and a simpler model is selected by measuring the complexity of the model, wherein the lower the AIC (akabane information content) and BIC (bayesian information content) values, the better the values, the lower the values, and the simpler the model.
Because the trend partial data extracted from the original sequence is predicted according to the obtained steps, the periodic characteristic sequence S in the final result needs to be addedi。
After the periodic component sequence obtained by preprocessing is added, the prediction formula of the ARIMA model is expressed as,
and S3, inputting the set future time length of the index variable to be predicted and the selected stable sequence which passes the d-order difference into the ARIMA model together, and calculating to obtain a target predicted value.
In a specific embodiment, the predicted future length of time is predicted as the predicted value of L': namely, setting the predicted value to beThe time definition data of the input required for prediction is expressed as L*={L1,L2,…LL+1,…LL+L′From L*、T′The vector can be obtained by repeated iterative computation of the input modelThe target prediction value of (1).
In a preferred embodiment, the method for predicting network metrics based on ARIMA model further includes:
and S4, overlapping the target predicted value and the periodic component sequence to obtain an actual predicted value.
It can be understood that, given a target time interval length L ' to be predicted, a stationary sequence T ' which is subjected to d-order difference is selected and used as a model input after training, and a target predicted value of the time interval length L ' is calculatedAnd predict the targetAnd adding the periodic characteristics to obtain a final actual predicted value.
In a preferred embodiment, the method for predicting network metrics based on ARIMA model further includes:
and S5, carrying out error calculation on the target predicted value and the actual predicted value to obtain a standard deviation and a relative error.
Aiming at the model prediction result, the error analysis is carried out on the predicted value and the actual value by respectively adopting two measurement indexes of standard deviation and relative error, wherein the standard deviationRelative error RMSE, smaller values of MAPE indicate more optimal prediction results.
In the actual operation process, a network optimization engineer can define and input model data indexes with different region granularities and different time spans according to the prediction requirement to carry out model training and prediction analysis, and the model effectively predicts a certain period of time in the future, so that the advance planning in the network operation and maintenance process is facilitated to a great extent, and an effective resource allocation strategy is formulated.
The embodiment mainly builds a prediction model based on periodicity (seasonality) by using an ARIMA time series data mining intelligent algorithm based on historically accumulated wireless network communication index data, effectively discovers and summarizes the change rule of the user on the wireless network demand in the historical data from the historical data change, effectively predicts the change sequence value of a certain index in a certain time in the future, and provides more effective data reference for optimizing wireless network resource allocation and performance.
Please refer to fig. 4-5. Another embodiment of the present invention further provides an ARIMA model-based network index prediction apparatus, including:
the acquisition module 21 is configured to acquire index data of an index variable to be predicted in a certain time period, and use the index data as a training data set.
Wherein the training data set comprises an index data sequence Xtrain={x1,x2,…xLAnd time series L ═ L1,L2,…LL};
In a specific embodiment, data within a certain historical time period (granularity of 1 hour/1 day) of the network communication index variable may be selected as a training data set of the model.
For example, a certain city is taken as a unit, time series historical data of a plurality of index variables collected by a base station of the city is selected, and if the change condition of the telephone traffic in a future week of a certain area needs to be predicted, the telephone traffic indexes with the granularity of the whole network, network elements or cells as the area granularity and the granularity of 1 hour as the selection time are taken as training data respectively.
And the model construction module 22 is configured to construct an ARIMA model for network index prediction by using the preprocessed training data set as input.
Preferably, the ARIMA model-based network index prediction apparatus further includes a preprocessing module 24, configured to decompose the training data set into a trend part, a period part, and a residual part; in particular, the method comprises the following steps of,
Xtrain={T1,T2,…TL}+{S1,S2,…SL}+{R1,R2,…RL};
xi=Ti+Si+Ri,i=1,2,…,L;
wherein the training data set comprises an index data sequence Xtrain={x1,x2,…xLAnd time series L ═ L1,L2,…LL};Ti、Si、RiRespectively is a trend component, a period component and an error term which are decomposed from the index data sequence.
In a specific embodiment, the periodic feature extraction is performed on a sequence with periodic features according to different time granularities, where a common algorithm STL (secure-Trend decoding procedure on Loess) in time sequence periodic rule computation decomposition is used, and data x at a certain time is extracted based on the LOESSiDecomposed into a trend component (T), a periodic component (S) and an error term (R).
Note that the trend component (T)i) The method is characterized by comprising a trend part; periodic component (S)i) Embodied by a periodic portion; error term (R)i) Embodying the residual part.
The building of the ARIMA model for predicting the network index refers to determining parameters p, d, q and other related parameters of the ARIMA model.
Determining parameters p, d, q specifically includes:
performing stationarity test and d-order difference processing on the trend component sequence obtained by preprocessing until a stationarity sequence is obtained, and determining a parameter d;
after the stationary sequence is obtained, the parameter p and the parameter q are determined by the PACF method and the ACF method, respectively.
In a specific embodiment, after the step periodic rule processing, the trend component sequence represents the original data sequenceBecause the trend component sequence T ═ T is taken1,T2,…TLAnd fourthly, carrying out next model parameter training.
And (3) carrying out statistical testing on stationarity: adopting ADF (automatic-Dickey-filler test) method to determine T as { T ═ T1,T2,…TLThe stationarity of the sequence is checked by unit root, if ADF passes, i.e. the sequence T '═ T'1,T′2,…T′LThe sequences are plateau sequences.
If the ADF check fails, i.e., the sequence T is a non-stationary sequence (there is a root of a unit), the sequence needs to be first order differenced, i.e.,
Ti_diff={Ti2-Ti1,Ti3-Ti2,…Tin-Tin-1}。
generally, the trend component sequence satisfies the ADF inspection after passing through the first order difference, and reaches a smooth sequence. However, if the first-order difference still fails to pass the ADF inspection, the cyclic difference processing is performed until the sequence stationary T ' ═ T ' is obtained '1,T′2,…T′LAnd d (the order of the cyclic difference).
Determining a parameter p: to obtain T '═ T'1,T′2,…T′LAfter the smoothing sequence, for the autoregressive process ar (p), the Partial Autocorrelation Coefficient (PACF) method is used to represent:
Φ11=ρ1;
wherein, using phikjRepresents the jth regression coefficient, rho, in the k-th order autoregressive equationjIs an autocorrelation coefficient, is obtained by the least square method, phikj=Φk-1,j-ΦkkΦk-1,k-1J ═ 1,2,3, …, k-1; when k is less than or equal to pTime phikkNot equal to 0; when k is>p is phikkApproaching 0, i.e. T'LAnd T'L-kThe partial autocorrelation coefficient of (a) is close to 0, which shows that the truncation characteristics of the partial autocorrelation function after the lag period p are effective, so that the order of the AR (p) process can be identified by using the characteristics to obtain a model parameter p.
Determining a parameter q: to obtain T '═ T'1,T′2,…T′LAfter the smoothing sequence, since the moving average method aims to effectively eliminate random fluctuation in prediction, for the moving average model process ma (q), the autocorrelation function (ACF) method is used to represent:
wherein, L is the sample capacity of the test sequence,is the sequence mean. According to the stable characteristic of the T' sequence, P can be judgedkIs a convergent sequence, i.e. when k>q is, PqApproaching 0 to determine the model parameters q.
Determining other related parameters specifically comprises:
after determining the parameters p, d, q, the prediction formula of the ARIMA model is expressed as:
where μ is a constant term, ρjIs the coefficient of an autoregressive process of order p, epsiloniIs an error constant, θjIs the error term coefficient of the moving average process of the order q.
In a specific embodiment, when there are multiple sets of candidate parameters p, d, and q that can be selected, the model parameters are selected according to the AIC and BIC values, and a simpler model is selected by measuring the complexity of the model, wherein the lower the AIC (akabane information content) and BIC (bayesian information content) values, the better the values, the lower the values, and the simpler the model.
Because the trend partial data extracted from the original sequence is predicted according to the obtained steps, the periodic characteristic sequence S in the final result needs to be addedi。
After the periodic component sequence obtained by preprocessing is added, the prediction formula of the ARIMA model is expressed as,
and the prediction module 23 is configured to input the set future time length of the index variable to be predicted and the selected stable sequence after the d-order difference are passed into the ARIMA model together, and calculate to obtain a target prediction value.
In a specific embodiment, the predicted future length of time is predicted as the predicted value of L': namely, setting the predicted value to beThe time definition data of the input required for prediction is expressed as L*={L1,L2,…LL+1,…LL+L′From L*Repeated iterative calculation of T' input model can obtain vectorThe target prediction value of (1).
In a preferred embodiment, the ARIMA model-based network index prediction apparatus further includes:
the prediction module 23 is further configured to superimpose the target prediction value and the periodic component sequence to obtain an actual prediction value.
It can be understood that, given a target time interval length L ' to be predicted, a stationary sequence T ' which is subjected to d-order difference is selected and used as a model input after training, and a target predicted value of the time interval length L ' is calculatedAnd predict the targetAnd adding the periodic characteristics to obtain a final actual predicted value.
In a preferred embodiment, the ARIMA model-based network index prediction apparatus further includes:
and the error analysis module 25 is configured to perform error calculation on the target predicted value and the actual predicted value to obtain a standard deviation and a relative error.
Aiming at the model prediction result, the error analysis is carried out on the predicted value and the actual value by respectively adopting two measurement indexes of standard deviation and relative error, wherein the standard deviationRelative error RMSE, smaller values of MAPE indicate more optimal prediction results.
In the actual operation process, a network optimization engineer can define and input model data indexes with different region granularities and different time spans according to the prediction requirement to carry out model training and prediction analysis, and the model effectively predicts a certain period of time in the future, so that the advance planning in the network operation and maintenance process is facilitated to a great extent, and an effective resource allocation strategy is formulated.
The embodiment mainly builds a prediction model based on periodicity (seasonality) by using an ARIMA time series data mining intelligent algorithm based on historically accumulated wireless network communication index data, effectively discovers and summarizes the change rule of the user on the wireless network demand in the historical data from the historical data change, effectively predicts the change sequence value of a certain index in a certain time in the future, and provides more effective data reference for optimizing wireless network resource allocation and performance.
Yet another embodiment of the present invention further provides a computer-readable storage medium comprising a stored computer program, wherein the computer program, when running, controls a device on which the computer-readable storage medium is located to perform the ARIMA model-based network index prediction method as described above.
The foregoing is directed to the preferred embodiment of the present invention, and it is understood that various changes and modifications may be made by one skilled in the art without departing from the spirit of the invention, and it is intended that such changes and modifications be considered as within the scope of the invention.
It will be understood by those skilled in the art that all or part of the processes of the methods of the embodiments described above can be implemented by a computer program, which can be stored in a computer-readable storage medium, and when executed, can include the processes of the embodiments of the methods described above. The storage medium may be a magnetic disk, an optical disk, a Read-Only Memory (ROM), a Random Access Memory (RAM), or the like.
Claims (9)
1. A network index prediction method based on an ARIMA model is characterized by comprising the following steps:
acquiring index data of an index variable to be predicted in a certain time period as a training data set;
taking the training data set after preprocessing as input, and constructing an ARIMA model for predicting network indexes;
inputting the set future time length of the index variable to be predicted and the selected stable sequence which passes the d-order difference into the ARIMA model together, and calculating to obtain a target predicted value;
the building of the ARIMA model for predicting the network indexes refers to determining parameters p, d, q and other related parameters of the ARIMA model;
determining parameters p, d, q specifically includes:
performing stationarity test and d-order difference processing on the trend component sequence obtained by preprocessing until a stationarity sequence is obtained, and determining a parameter d;
after the stable sequence is obtained, determining a parameter p and a parameter q by respectively using a PACF method and an ACF method;
determining other related parameters specifically comprises:
after determining the parameters p, d, q, the prediction formula of the ARIMA model is expressed as:
where μ is a constant term, ρjIs the coefficient of an autoregressive process of order p, epsiloniIs an error constant, θjError term coefficients for a moving average process of order q;
after the periodic component sequence obtained by preprocessing is added, the prediction formula of the ARIMA model is expressed as,
2. the ARIMA model-based network metric prediction method of claim 1 wherein the training data set comprises an metric data sequence Xtrain={x1,x2,...xLAnd time series L ═ L1,L2,...LL};
The training data set is preprocessed by decomposing the training data set into a trend part, a period part and a residual part; in particular, the method comprises the following steps of,
Xtrain={T1,T2,...TL}+{S1,S2,...SL}+{R1,R2,...RL};
xi=Ti+Si+Ri,i=1,2,...,L;
wherein, Ti、Si、RiRespectively is a trend component, a period component and an error term which are decomposed from the index data sequence.
3. The ARIMA model-based network indicator prediction method of claim 1, further comprising:
and superposing the target predicted value and the periodic component sequence to obtain an actual predicted value.
4. The ARIMA model-based network indicator prediction method of claim 3, further comprising:
and carrying out error calculation on the target predicted value and the actual predicted value to obtain a standard deviation and a relative error.
5. A network index prediction device based on an ARIMA model is characterized by comprising:
the acquisition module is used for acquiring index data of an index variable to be predicted in a certain time period as a training data set;
the model construction module is used for taking the training data set after preprocessing as input to construct an ARIMA model for predicting network indexes;
the prediction module is used for inputting the set future time length of the index variable to be predicted and the selected stable sequence which passes the d-order difference into the ARIMA model together to calculate and obtain a target prediction value;
the building of the ARIMA model for predicting the network indexes refers to determining parameters p, d, q and other related parameters of the ARIMA model;
determining parameters p, d, q specifically includes:
performing stationarity test and d-order difference processing on the trend component sequence obtained by preprocessing until a stationarity sequence is obtained, and determining a parameter d;
after the stable sequence is obtained, determining a parameter p and a parameter q by respectively using a PACF method and an ACF method;
determining other related parameters specifically comprises:
after determining the parameters p, d, q, the prediction formula of the ARIMA model is expressed as:
where μ is a constant term, ρjIs the coefficient of an autoregressive process of order p, epsiloniIs an error constant, θjError term coefficients for a moving average process of order q;
after the periodic component sequence obtained by preprocessing is added, the prediction formula of the ARIMA model is expressed as,
6. the ARIMA model-based network metric prediction device of claim 5, further comprising a pre-processing module for decomposing the training data set into a trend part, a periodic part, and a residual part; in particular, the method comprises the following steps of,
Xtrain={T1,T2,...TL}+{S1,S2,...SL}+{R1,R2,...RL};
xi=Ti+Si+Ri,i=1,2,...,L;
wherein the training data set comprises an index data sequence Xtrain={x1,x2,...xLAnd time series L ═ L1,L2,...LL};Ti、Si、RiRespectively is a trend component, a period component and an error term which are decomposed from the index data sequence.
7. The ARIMA model-based network metric prediction device of claim 5, further comprising:
and the prediction module is also used for superposing the target predicted value and the periodic component sequence to obtain an actual predicted value.
8. The ARIMA model-based network metric prediction device of claim 7, further comprising:
and the error analysis module is used for carrying out error calculation on the target predicted value and the actual predicted value to obtain a standard deviation and a relative error.
9. A computer-readable storage medium storing a computer program, wherein the computer program when executed controls a device on which the computer-readable storage medium is located to perform the ARIMA model-based network metrics prediction method according to any of claims 1 to 4.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201811485502.2A CN109587713B (en) | 2018-12-05 | 2018-12-05 | Network index prediction method and device based on ARIMA model and storage medium |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201811485502.2A CN109587713B (en) | 2018-12-05 | 2018-12-05 | Network index prediction method and device based on ARIMA model and storage medium |
Publications (2)
Publication Number | Publication Date |
---|---|
CN109587713A CN109587713A (en) | 2019-04-05 |
CN109587713B true CN109587713B (en) | 2022-01-11 |
Family
ID=65926106
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201811485502.2A Active CN109587713B (en) | 2018-12-05 | 2018-12-05 | Network index prediction method and device based on ARIMA model and storage medium |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN109587713B (en) |
Families Citing this family (21)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110070215B (en) * | 2019-04-09 | 2022-08-23 | 南京邮电大学 | Solar energy collection power prediction method based on adaptive seasonal ARIMA model |
CN110062114A (en) * | 2019-04-09 | 2019-07-26 | 国家计算机网络与信息安全管理中心 | Fraudulent call prediction technique and forecasting system based on ARIMA |
CN110298690B (en) * | 2019-05-31 | 2023-07-18 | 创新先进技术有限公司 | Object class purpose period judging method, device, server and readable storage medium |
CN112149860A (en) * | 2019-06-28 | 2020-12-29 | 中国电力科学研究院有限公司 | Automatic anomaly detection method and system |
CN112203320B (en) * | 2019-07-08 | 2023-04-28 | 中国移动通信集团贵州有限公司 | Method and device for predicting target network parameters based on gray model |
CN110674019B (en) * | 2019-08-30 | 2023-02-07 | 中国人民财产保险股份有限公司 | Method and device for predicting system fault and electronic equipment |
CN113114520A (en) * | 2020-01-10 | 2021-07-13 | 华为技术有限公司 | Data transmission method and terminal |
CN111241066B (en) * | 2020-01-10 | 2024-06-25 | 平安科技(深圳)有限公司 | Platform database automation operation and maintenance method, device and computer readable storage medium |
CN111444246B (en) * | 2020-05-07 | 2021-01-29 | 北京工业大数据创新中心有限公司 | Granularity linear trend automatic generation method and system based on time series data |
CN111611519B (en) * | 2020-05-28 | 2023-07-11 | 上海观安信息技术股份有限公司 | Method and device for detecting personal abnormal behaviors |
CN112115416A (en) * | 2020-08-06 | 2020-12-22 | 深圳市水务科技有限公司 | Predictive maintenance method, apparatus, and storage medium |
CN112131212A (en) * | 2020-09-29 | 2020-12-25 | 合肥城市云数据中心股份有限公司 | Hybrid cloud scene-oriented time sequence data anomaly prediction method based on ensemble learning technology |
CN111898758B (en) * | 2020-09-29 | 2021-03-02 | 苏宁金融科技(南京)有限公司 | User abnormal behavior identification method and device and computer readable storage medium |
CN114428803B (en) * | 2020-10-29 | 2023-05-26 | 上海浦昊节能环保科技有限公司 | Air compression station operation optimization method, system, storage medium and terminal |
CN112862159B (en) * | 2021-01-13 | 2023-04-28 | 中铁第四勘察设计院集团有限公司 | Method, device, equipment and storage medium for predicting total electron content of ionized layer |
CN112884503A (en) * | 2021-01-21 | 2021-06-01 | 百果园技术(新加坡)有限公司 | User scale prediction method, device, equipment and medium |
CN112929214A (en) * | 2021-02-02 | 2021-06-08 | 北京明朝万达科技股份有限公司 | Model construction method, device, equipment and storage medium |
CN112929215B (en) * | 2021-02-04 | 2022-10-21 | 博瑞得科技有限公司 | Network flow prediction method, system, computer equipment and storage medium |
CN113329128B (en) * | 2021-06-04 | 2023-01-06 | 中国电信股份有限公司 | Traffic data prediction method and device, electronic equipment and storage medium |
CN113672642A (en) * | 2021-07-07 | 2021-11-19 | 德特赛维技术有限公司 | Method and system for predicting mobile network performance based on time sequence |
CN117455666A (en) * | 2023-10-16 | 2024-01-26 | 厦门国际银行股份有限公司 | Transaction technical index prediction method, device and equipment based on neural network |
Family Cites Families (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102088709A (en) * | 2010-11-30 | 2011-06-08 | 哈尔滨工业大学 | Method for predicting telephone traffic based on clustering and autoregressive integrated moving average (ARIMA) model |
CN104994539B (en) * | 2015-06-30 | 2018-03-30 | 电子科技大学 | A kind of wireless sensor network Traffic anomaly detection method based on ARIMA models |
-
2018
- 2018-12-05 CN CN201811485502.2A patent/CN109587713B/en active Active
Also Published As
Publication number | Publication date |
---|---|
CN109587713A (en) | 2019-04-05 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN109587713B (en) | Network index prediction method and device based on ARIMA model and storage medium | |
CN109886430B (en) | Equipment health state assessment and prediction method based on industrial big data | |
Verenich et al. | Survey and cross-benchmark comparison of remaining time prediction methods in business process monitoring | |
CN112910690A (en) | Network traffic prediction method, device and equipment based on neural network model | |
CN110930198A (en) | Electric energy substitution potential prediction method and system based on random forest, storage medium and computer equipment | |
CN110621026B (en) | Multi-moment prediction method for base station flow | |
CN109376906B (en) | Travel time prediction method and system based on multi-dimensional trajectory and electronic equipment | |
CN114358445A (en) | Business process residual time prediction model recommendation method and system | |
CN115394383A (en) | Method and system for predicting luminescence wavelength of phosphorescent material | |
CN113704389A (en) | Data evaluation method and device, computer equipment and storage medium | |
CN115545331A (en) | Control strategy prediction method and device, equipment and storage medium | |
CN108595526A (en) | Resource recommendation method and device | |
CN109255389B (en) | Equipment evaluation method, device, equipment and readable storage medium | |
CN110516792A (en) | Non-stable time series forecasting method based on wavelet decomposition and shallow-layer neural network | |
CN115883424B (en) | Method and system for predicting flow data between high-speed backbone networks | |
CN113762356B (en) | Cluster load prediction method and system based on clustering and attention mechanism | |
CN115794548A (en) | Method and device for detecting log abnormity | |
CN111654853B (en) | Data analysis method based on user information | |
CN114330720A (en) | Knowledge graph construction method and device for cloud computing and storage medium | |
CN113868462A (en) | Song recommendation system and method based on matrix decomposition | |
da Silva et al. | Automated machine learning for time series prediction | |
CN117131999B (en) | Digital twin-based rail transit passenger flow prediction system and method thereof | |
CN115237606B (en) | Edge calculation control system | |
CN118627645A (en) | Long-time sequence prediction integrated learning device based on multicycle divergence | |
CN117009752A (en) | Power consumption demand prediction method based on POA-VMD-SSA-KELM model |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |